JP2023087699A - Level estimation method for face showing actual feeling of happiness, level estimation device for face showing actual feeling of happiness, and level estimation program for face showing actual feeling of happiness - Google Patents

Abstract

To provide a novel estimation method, an estimation device and an estimation program that estimate a level of "a face showing an actual feeling of happiness" with high precision and good reproducibility.SOLUTION: An estimation method includes estimating a level of the face of an object person showing an actual feeling of happiness from face image data of the object person based upon correlation between the intensity of a spatial frequency obtained from the face image data and an evaluated value of the face showing the actual feeling of happiness.EFFECT: Consequently, a novel technique can be provided which estimates the level of "the face showing the actual feeling of happiness" with high precision and good reproducibility.SELECTED DRAWING: None

Description

The present invention relates to a method for estimating the degree of happy faces, a device for estimating the degree of happy faces, and a program for estimating the degree of happy faces.

It is one of the important evaluations for everyone to have a "face that looks happy/a face that expresses happiness". In particular, the evaluation of the degree of "a face that looks happy/a face that expresses a sense of happiness" is extremely important in various fields such as esthetics and cosmetics.

2. Description of the Related Art Conventionally, a monitoring/recognition system is known in which a computer captures a face image, recognizes age group, gender, and facial characteristics, and estimates a person. Patent Documents 1 to 3 describe an age estimation system using image processing values such as edging and binarization (for example, Patent Documents 1 and 2), feature amounts of feature points such as eyes, nose and mouth (position, shading, etc.). , the number of wrinkles) is disclosed (for example, Patent Document 3).

There is also known a technique that focuses on the fact that the texture characteristics of an image change depending on spatial periodicity (spatial frequency). As a technique focusing on spatial frequency, Patent Document 4 discloses a method of creating a simulation image of skin having various textures by changing spatial frequency.

In addition, the applicant of the present application has disclosed a technique for analyzing the pattern of spatial frequencies during makeup and evaluating the effect of makeup on the "three-dimensional effect of the face" (for example, Patent Document 5).

JP 2009-086901 A JP 2009-271885 A Japanese Patent Application Laid-Open No. 2008-282089 JP-A-2004-283357 JP 2004-272849 A

Where there is the above prior art, as a result of diligent research efforts by the present inventors, there is a high correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing happiness. I found
That is, an object of the present invention is to provide a novel technique for estimating the degree of "a face expressing happiness" with high accuracy and good reproducibility.

The present invention for solving the above-mentioned problems is based on the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing the feeling of happiness. It is an estimation method that includes estimating the degree of face expression.
As shown in the examples described later, the intensity of the spatial frequency has a high correlation with a happy face.
Therefore, according to the estimation method of the present invention, it is possible to estimate, from the face image data of the subject with high accuracy and good reproducibility, the degree to which the subject's feeling of happiness appears.

A preferred embodiment of the present invention is a spatial frequency calculation step of calculating the intensity of spatial frequencies obtained by combining data obtained from arbitrary channels among RGB and BW color channel data of color face image data. and,
Applying the calculated spatial frequency intensity to a previously prepared formula or model showing the correlation between the spatial frequency intensity obtained from the face image data and the evaluation value of the face expressing happiness, an evaluation process for evaluating the degree of happiness of the face,
is an estimation method including

A preferred embodiment of the present invention is characterized in that the formula is a multiple regression formula obtained by multiple regression analysis or a prediction formula obtained by PLS.
By using a multiple regression equation or a prediction equation, it is possible to estimate the degree of the subject's feeling of happiness from the face image data of the subject with high accuracy and good reproducibility.

In a preferred embodiment of the present invention, the degree of the face expressing happiness is the degree of lift-up of the cheeks, the stereoscopic effect around the mouth, the uniformity of the skin color, the smoothness of the skin, and the brightness around the eyes. Including at least some degree.

In a preferred embodiment of the present invention, the face image data is face image data of a part of the face or the whole face.

In a preferred embodiment of the present invention, the face image data is color face image data, and the spatial frequency is a spatial frequency obtained from at least one of RGB and BW color channel data of the color face image data.

In a preferred embodiment of the present invention, the spatial frequency is a spatial frequency obtained from a plurality of color channel data among RGB and BW color channel data of the face image.
Of the RGB and BW color channel data of the face image, by combining and using the intensity of the spatial frequency obtained from a plurality of color channel data, the happiness of the subject can be obtained from the face image data of the subject with high accuracy and good reproducibility. It is possible to estimate (evaluate) the degree to which the actual feeling is expressed on the face.

In a preferred form of the invention, the spatial frequency range (cycle/image-width) is from 0 to 512 cycles/image-width for an image with an image length of 1024 pixels.

In addition, the present invention provides an application step of applying any stimulus (excluding medical practice) to the subject;
a measuring step of measuring the degree of happiness-appearing faces obtained before the applying step and estimated by the above-described estimation method, and the degree of change in the degree of happiness-appearing faces after the applying step;
a determination step of determining that the stimulus is suitable for the subject when an increase in the degree of happiness is observed on the face, based on the result of the measurement step;
It is also a method of selecting a stimulus suitable for a subject, characterized by having

Further, the present invention provides an application step of applying a skin condition improving agent and/or cosmetic means to a subject,
a measuring step of measuring the degree of happiness-appearing faces obtained before the applying step and estimated by the above-described estimation method, and the degree of change in the degree of happiness-appearing faces after the applying step;
a determination step of determining that the skin condition improving agent and/or cosmetic means are more suitable for the subject as the rate of increase in the degree of happiness manifestation increases based on the results of the measuring step;
A method for selecting a skin condition improving agent and/or cosmetic means, characterized by having

The present invention is also a screening method for a skin condition improving agent based on the correlation between the intensity of spatial frequency obtained from face image data and the evaluation value of a face expressing happiness.

In a preferred embodiment of the method for screening a skin condition improving agent of the present invention, an applying step of applying a substance to the skin;
a measuring step of measuring the degree of happiness-appearing faces obtained before the applying step and estimated by the above-described estimation method, and the degree of change in the degree of happiness-appearing faces after the applying step;
A determination step of determining that a skin condition improving agent with a higher rate of increase in the degree of happiness expressed on the face is a superior skin condition improving agent based on the results of the measuring step;
characterized by having

In addition, the present invention for solving the above problems calculates the intensity of the spatial frequency obtained by combining the data obtained from arbitrary channels among the RGB and BW color channel data of the color face image data. spatial frequency calculation means;
The calculated intensity of spatial frequency is applied to a previously prepared formula or model showing the correlation between the intensity of spatial frequency obtained from face image data and a face expressing happiness, so that the subject feels happiness. an evaluation means for evaluating the degree of the expression of the face;
An estimating device comprising:
According to the estimation device of the present invention, it is possible to estimate the degree of the subject's feeling of happiness from the face image data of the subject with high accuracy and good reproducibility.

Further, the present invention for solving the above problems provides a computer,
Spatial frequency calculation means for calculating the strength of spatial frequencies obtained by combining data obtained from arbitrary channels among RGB and BW color channel data of color face image data, and
The calculated intensity of spatial frequency is applied to a previously prepared formula or model showing the correlation between the intensity of spatial frequency obtained from face image data and a face expressing happiness, so that the subject feels happiness. an evaluation means for evaluating the degree of the expression of the face;
It relates to an estimation program for functioning as
According to the estimation program of the present invention, a computer can estimate the degree of the subject's feeling of happiness from the face image data of the subject with high accuracy and good reproducibility. .

According to the present invention, it is possible to provide a novel technique for estimating the degree of "a face expressing happiness" with high accuracy and good reproducibility.

FIG. 4 is a diagram showing representative examples of a “face that looks happy” and a “face that does not look happy”. FIG. 10 is a diagram showing the relationship between the average evaluation value (hereinafter also referred to as score value) of “a face that seems to have a feeling of happiness” and age. FIG. 4 is a diagram showing correlation coefficients between score values and power values of respective spatial frequencies; FIG. 10 is a diagram showing the relationship between estimated values of multiple regression analysis results and score values of faces that appear to have a sense of happiness. It is an image in which the power value of R4 c/iw has been removed (0-fold processing) or enhanced (5-fold processing). It is an image in which the power value of R6 c/iw has been removed (0x processing) or emphasized (5x processing). It is an image in which the power value of R10 c/iw has been removed (0x processing) or emphasized (5x processing). It is an image in which the power value of R11 c/iw has been removed (0x processing) or emphasized (5x processing). It is an image in which the power value of R12 c/iw has been removed (0x processing) or emphasized (5x processing). It is an image in which the power value of R13 c/iw has been removed (0x processing) or emphasized (5x processing). It is an image in which the power values of R24 c/iw and R29 c/iw are removed (0-fold processing) or emphasized (5-fold processing). Remove the power values of R68 c/iw, R79 c/iw, R80 c/iw, R88 c/iw, R109 c/iw, R112 c/iw, R127 c/iw, R133 c/iw (0x processing) or It is an image that has been enhanced (10 times processed). This is an image in which the power value of G5 c/iw has been removed (0-fold processing) or emphasized (5-fold processing). It is an image in which the power value of G8 c/iw has been removed (0-fold processing) or emphasized (5-fold processing). It is an image in which the power value of G12 c/iw has been removed (0-fold processing) or enhanced (5-fold processing). This is an image in which the power value of G13 c/iw has been removed (0x processing) or emphasized (5x processing). It is an image in which the power values of G22 c/iw and G26 c/iw are removed (0-fold processing) or emphasized (5-fold processing). This is an image in which the power values of G33 c/iw, G105 c/iw, G115 c/iw, G125 c/iw, and G132 c/iw are removed (0x processing) or enhanced (10x processing). It is an image in which the power values of G43 c/iw, G110 c/iw, and G128 c/iw are removed (0-fold processing) or emphasized (10-fold processing). B9 This is an image in which the power value of c/iw has been removed (0x processing) or emphasized (5x processing). It is an image in which the power value of B15 c/iw has been removed (0x processing) or emphasized (5x processing). It is an image in which the power values of B28 c/iw and B47 c/iw are removed (0-fold processing) or emphasized (10-fold processing). B36 c/iw, B41 c/iw, B48 c/iw, B57 c/iw, B68 c/iw, B93 c/iw, B102 c/iw, B108 c/iw, B122 c/iw power values removed ( 0-fold processing) or enhanced (10-fold processing). It is an image in which the power value of BW2 c/iw has been removed (0x processing) or emphasized (5x processing). These are images in which the power values of BW46 c/iw, BW63 c/iw, BW75 c/iw, and BW85 c/iw have been removed (0-fold processing) or emphasized (10-fold processing). It is an image in which the power values of BW74 c/iw and BW83 c/iw are removed (0-fold processing) or emphasized (10-fold processing). FIG. 4 is a flow diagram showing a specific aspect of the estimation method of the present invention; It is a hardware block diagram of an estimation device of the present invention.

<Method for estimating the degree of happiness expressed on the face>
According to the present invention, a face showing a feeling of happiness of a subject is identified from the face image data of the subject based on the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face showing the feeling of happiness. Including estimating the degree.

The "face expressing happiness" referred to in the present invention refers to a face evaluated as a face expressing happiness based on the objective impression of a third party.
Specifically, the ``face expressing happiness'' as used in the present invention means ``a person who appears to be feeling happiness'' even though he/she has a straight face. It refers to a face that is evaluated as a face that expresses a sense of happiness based on an objective impression.

The ``estimation of the degree of happiness on the face'' according to the present invention includes the degree of lift-up of the cheeks, the three-dimensional effect around the mouth, the uniformity of the skin color, the smoothness of the skin, Brightness around the eyes, preferably 2 or more, more preferably 3 or more, even more preferably 4 or more, and particularly preferably all degrees.

Acquisition of "face image data" can be performed according to a standard method. As a method of acquiring the "face image data", for example, a method of taking an image of a front face using a digital camera or the like and importing the imaged information into a personal computer as digital information can be mentioned.

As the face image data of the subject used in the present invention, face image data of the entire face or parts of the face can be used.

When face image data of the whole face is used, it is more preferable to use the whole face data excluding the hair portion. As a result, it is possible to estimate, from the face image data of the subject, the extent to which the subject's feeling of happiness is manifested with high accuracy and good reproducibility.
In this case, the size of the face image is in the range of about 30 cm × 30 cm, considering the sampling theorem for calculating the resolution and spatial frequency intensity necessary for estimating the degree of happiness of the face. It is preferable that the image is an image of 512×512 pixels or more.

Face image data for facial parts include face image data for cheeks, forehead, chin, nose bridge, upper corners of eyes (T-shaped eyes and nose), under eyes, lips, eyes, nose, forehead, cheeks, neck, etc. is mentioned.
Among them, facial image data of the entire cheek, the entire forehead, the entire chin, the bridge of the nose, the upper part of the outer corner of the eye (T-shape of the eyes and nose) and/or the lower part of the eye can be preferably used.

As shown in the examples described later, the intensity of the spatial frequency obtained from the facial image data of the entire cheek, the entire forehead, the entire chin, the bridge of the nose, the upper part of the outer corner of the eye (T-shaped eye and nose) and/or the lower eye is an expression of happiness. It is highly correlated with the face.
Therefore, by using facial image data of the entire cheek, the entire forehead, the entire chin, the bridge of the nose, the upper part of the corner of the eye (T shape of the eyes and nose) and/or the lower part of the eye, the facial image data of the subject can be obtained with high accuracy and good reproducibility. It is possible to estimate the degree to which the subject's face expresses happiness.

When using images of the entire cheeks, the entire forehead, the entire chin, the bridge of the nose, the upper parts of the corners of the eyes (T-shaped eyes and nose), and/or the lower parts of the eyes, it is possible to use images of at least these areas of about 2 cm x 2 cm. preferable. By acquiring information on spatial frequency obtained from face image data of the entire cheek, the entire forehead, the entire chin, the bridge of the nose, the upper part of the corner of the eye (T-shaped eye and nose) and / or the face under the eye, it is possible to achieve high accuracy and reproducibility. The degree of the subject's feeling of happiness can be estimated from the face image data. In these cases, the number of pixels of face image data is preferably 256 pixels or more, more preferably 512 pixels or more.

Preferably, the present invention uses color channel data of any one of R (red), G (green), B (blue), and BW (gray) of the face image for estimation.

In this specification, "color channel data of any one of R (red), G (green), B (blue), and BW (gray) of a face image" means that a color face image is divided into RGB and BW color channels. color channel data that retains a specific hue from among the face image data obtained.

As a method of acquiring color channel data of any one of R (red), G (green), B (blue), and BW (gray) of a face image, a color face image is provided to an image processing system, and color channel division is performed. You can mention how to do it.

By using any one of R (red), G (green), B (blue), and BW (gray) color channel data of the face image, the face image data of the subject can be obtained from the face image data with high accuracy and good reproducibility. It is possible to estimate the degree of happiness expressed on the face of each person.

In the present invention, the spatial frequency is preferably obtained by combining two or more, more preferably three types of color channel data among the RGB and BW color channel data of the face image. is preferred.
By using spatial frequency intensities obtained from multiple color channel data out of the RGB and BW color channel data of the face image, it is possible to estimate with high accuracy and good reproducibility the degree to which the subject's face expresses a sense of happiness. can do.

"Spatial frequency" is a property of a structure that has spatial periodicity and refers to the number of repetitions of a structure contained in a unit length. Here, "spatial frequency" refers to the period per meter in the International System of Units, but the number of lines per millimeter is used in the field of image processing.
Moreover, the number of cycles per image length (cycle/image-width) can also be used as the unit of spatial frequency.
The maximum number of cycles included in an image is 1/2 the number of pixels (pixels, dots) included in the length of the image. That is, an image with an image length of 512 pixels has a spatial frequency band of 0 to 256 cycles/image-width, and an image with an image length of 1024 pixels has a spatial frequency band of 0 to 512 cycles/image-width.

In this specification, when referring to the spatial frequency range of any one of R (red), G (green), B (blue), and BW (gray) color channel data of a face image, it is expressed as "hue spatial frequency range". do.
For example, of the R (red), G (green), B (blue), and BW (gray) color channel data of the face image, the spatial frequency range obtained from the R (red) color channel data is 4 c/ If it is iw, it is expressed as "R 4 c/iw". (See Table 1 Fig. 5).

"Intensity of spatial frequency" is an amplitude value or power (power logarithm) at each spatial frequency. Spatial frequency intensities can be calculated according to common methods.

Specifically, the face image data obtained by the above-described method is subjected to color channel division into either RGB or BW color channels, and then commercially available image analysis software (for example, Fourier phase analysis program manufactured by ImageSense Inc.). or PoplmaGinG (manufactured by Digital Being Kids, Inc.) or the like to perform FFT (FastFouRieRTRansfoRm: Fast Fourier Transform), the amplitude value or power (intensity of spatial frequency) can be calculated.

As an example of spatial frequency analysis, FIG. 2 shows a spectrum pattern resulting from FFT processing of a gray image of 1024×1024 pixels (in the case of color face image data, an image divided into color channels).

FFT processing can calculate the intensity (amplitude value or power value) for each spatial frequency (unit: cycle/image-width) from 0 to 1024/2.
The calculated intensity (amplitude value or power value) of each spatial frequency reveals the spatial frequency characteristics of the face image.

A great advantage of the present invention is that numerical data (spatial frequency intensity (power spectrum)) necessary for estimating the degree of happiness manifested in a subject's face image data can be easily obtained. In the present invention, it can be easily and quantitatively calculated using commercially available software.

The ``estimation of the degree of happiness expressed by the face'' in the present invention may be a quantitative estimate, or a relative and qualitative estimation based on a preset ``index of the face expressing happiness''. good.

Specifically, as a quantitative estimation method, a formula or model showing the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing happiness is created in advance. By comparing the intensity of the spatial frequency obtained from any of the color channel data of R (red), G (green), B (blue), and BW (gray) obtained from the formula or model, the subject's A preferable example is an embodiment in which the degree of happiness expressed by the face is estimated. By adopting such an embodiment, it is possible to improve the estimation accuracy.

Here, the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing happiness can be obtained in advance by multivariate analysis such as regression analysis. The multivariate analysis is preferably one that can utilize the relationship between the objective variable and the explanatory variable, and can preferably be exemplified by discriminant analysis and regression analysis (MLR, PLS, PCR, logistic).

Among these, particularly preferred are multiple regression analysis (MLR) and nonlinear regression analysis (PLS: PaRtialLeastSquaRes). For example, a multiple regression equation can be obtained by performing multiple regression analysis using the spatial frequency intensity as an explanatory variable and a happy face as an objective variable. Also, a prediction formula (prediction model) can be obtained by performing PLS in the same manner.

A method of preparing an equation or a model showing the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing happiness will be described below.

Principal component analysis, factor analysis, quantification theory 1, quantification theory 2, quantification theory 3, multidimensional scaling, supervised clustering, neural networks, ensemble learning, Multivariate analysis such as can be used as appropriate. Preferred among these are neural networks, discriminant analysis and quantification theory. These multivariate analyzes can be performed using free software or commercially available software.

Here, in order to obtain a highly accurate formula or model, the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing happiness that is visually evaluated in advance is stored as input information. a storage unit;
a learning unit that generates a learned model in which a variable for estimating the degree of the subject's face showing a sense of happiness is machine-learned from the subject's face image data using the correlation as teacher data;
A learning device can also be used, comprising:

In order to create a formula or model showing the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing happiness, the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing happiness It is preferable to create a database (DB) in which correlations with face evaluation values are associated. Here, the number of DBs is preferably 10 or more, more preferably 20 or more, more preferably 50 or more, more preferably 100 or more, still more preferably 200 or more.
As for the structure of the DB, for example, if it is a matrix format, the strength of the spatial frequency obtained from the face image data can be entered in the row, and the degree of happiness expressed in the face can be entered in the column.

This DB is updated by adding the estimated value after estimating the correlation between the intensity of the spatial frequency obtained from the newly acquired face image data of the subject and the evaluation value of the face expressing happiness. You may If desired, the multivariate analysis described above can be performed on the updated DB to update the formulas or models.

A method of obtaining information necessary for creating a formula or a model or creating the above DB will be described below.

(1) Evaluation of faces expressing happiness , visually evaluate the faces expressing happiness for a plurality of persons. The number of persons evaluated is at least 50 or more, preferably 100 or more, more preferably 500 or more. In addition, it is preferable that the population distribution is approximately the same for each age group. Since there is often a gap between the evaluation by the person himself and the evaluation by a third party with respect to the face expressing happiness, it is desirable that the evaluator be a third party. The selection of evaluators can be done from the viewpoint of selecting an appropriate evaluator to represent a third party. etc., it is important to consider various issues.
For example, an evaluator having experience and expertise in evaluation in the field of beauty is preferable. The number of evaluators is preferably plural, specifically about 5 to 10 persons. It is preferable to statistically process evaluation results by a plurality of evaluators.

Also, an average value, a median value, or the like, excluding outliers, is calculated, and the degree of the face expressing objective happiness of each person can be obtained.
It is desirable to evaluate the degree of happiness expressed by a face using a face image, as shown in Examples described later. This is because it is possible to exclude the effects of lowering the evaluation accuracy due to facial expressions, hairstyles, backgrounds, skin colors, skin troubles, and the like.

(2) Calculation of Intensity of Spatial Frequency Next, the intensity of spatial frequency is calculated from the face image data showing the plurality of persons. Calculation of the intensity of the spatial frequency is as described above. The face image used here is preferably the same as the face image used in the evaluation of the face expressing happiness. This makes it possible to more accurately obtain the correlation between the intensity of the spatial frequency and the evaluation value of the face expressing happiness.

A preferred embodiment of the spatial frequency used in the estimation method of the present invention will be described in detail below.

The estimation method of the present invention includes a spatial frequency (cycle/image-width) of a value of preferably 1/5 or less, more preferably 1/10, and even more preferably 1/15 of the image length (pixel). is preferably used.

In the following description, unless otherwise specified, it refers to a preferred form when face image data with an image length of 1024 pixels is used.

In addition, the estimation method of the present invention preferably includes 2 or more, more preferably 3 or more, more preferably 5 or more, more preferably 8 or more, still more preferably 10 or more, still more preferably 15 or more, particularly preferably all Preferably, the intensity of the spatial frequency of the spatial frequency is used.

The estimation method of the present invention includes
R4 c/iw, R6 c/iw, R10 c/iw, R11 c/iw, R24 c/iw, R29 c/iw, G5 c/iw, G8 c/iw, G13 c/iw, G22 c/iw, from G26 c/iw, B9 c/iw, B15 c/iw, B28 c/iw, B47 c/iw, BW2 c/iw, BW46 c/iw, BW63 c/iw, BW75 c/iw, BW85 c/iw 1 or more selected, preferably 2 or more, more preferably 3 or more, more preferably 5 or more, more preferably 8 or more, still more preferably 10 or more, still more preferably 15 or more, particularly preferably at all spatial frequencies It is preferred to use the intensity of the frequency.

In a particularly preferred embodiment of the invention, R4 c/iw, R6 c/iw, R10 c/iw, R11 c/iw, R24 c/iw, R29 c/iw, G5 c/iw, G8 c/iw, G13 c/iw, G22 c/iw, G26 c/iw, B9 c/iw, B15 c/iw, B28 c/iw, B47 c/iw, BW2 c/iw, BW46 c/iw, BW63 c/iw, The intensity of the spatial frequency selected from BW75 c/iw and BW85 c/iw is remarkable (when the sign (positive or negative) of the correlation coefficient shown in Table 1 and FIG. 3 is positive, the higher the intensity, the more If the sign (positive or negative) of the correlation coefficient shown in 3 is negative, the intensity is low), the degree of cheek lift-up is judged to be excellent, and based on the judgment, it is estimated that the face is a face that expresses a sense of happiness. do.

In addition, the estimation method of the present invention includes R4 c/iw, R11 c/iw, R12 c/iw, R13 c/iw, R24 c/iw, R29 c/iw, G5 c/iw, G8 c/iw, 1 or more selected from G12 c/iw, G13 c/iw, B9 c/iw, B15 c/iw, BW2 c/iw, BW46 c/iw, BW63 c/iw, BW75 c/iw, BW85 c/iw , preferably 2 or more, more preferably 3 or more, more preferably 5 or more, more preferably 8 or more, still more preferably 10 or more, and particularly preferably all spatial frequencies.

In a particularly preferred embodiment of the invention, R4 c/iw, R11 c/iw, R12 c/iw, R13 c/iw, R24 c/iw, R29 c/iw, G5 c/iw, G8 c/iw, of spatial frequencies selected from G12 c/iw, G13 c/iw, B9 c/iw, B15 c/iw, BW2 c/iw, BW46 c/iw, BW63 c/iw, BW75 c/iw, and BW85 c/iw The intensity is remarkable (when the sign (positive or negative) of the correlation coefficient shown in Table 1 and FIG. 3 is positive, the higher the intensity, the more intensity is lower), it is determined that the three-dimensional effect around the mouth is excellent, and based on the determination, it is estimated that the face expresses a sense of happiness.

In addition, the estimation method of the present invention includes R10 c/iw, G5 c/iw, G8 c/iw, G12 c/iw, G22 c/iw, G26 c/iw, B9 c/iw, B15 c/iw, B28 c / iw, B47 c / iw, selected from 1 or more, preferably 2 or more, more preferably 3 or more, more preferably 5 or more, more preferably 8 or more, still more preferably 10 or more, particularly preferably all It is preferred to use the intensity of the spatial frequency at the spatial frequency.

In a particularly preferred embodiment of the invention, R10 c/iw, G5 c/iw, G8 c/iw, G12 c/iw, G22 c/iw, G26 c/iw, B9 c/iw, B15 c/iw, The strength of the spatial frequency selected from B28 c/iw and B47 c/iw is remarkable (when the sign (positive or negative) of the correlation coefficient shown in Table 1 and FIG. 3 is positive, the higher the strength, the higher the If the sign (positive/negative) of the correlation coefficient shown in 3 is negative, the strength is low), it is determined that the uniformity of the skin color is excellent, and based on the determination, it is estimated that the face expresses a sense of happiness.

In addition, the estimation method of the present invention includes R68 c/iw, R79 c/iw, R80 c/iw, R88 c/iw, R109 c/iw, R112 c/iw, R127 c/iw, R133 c/iw, G33 c/iw, G105 c/iw, G115 c/iw, G125 c/iw, G132 c/iw, G43 c/iw, G110 c/iw, G128 c/iw, B36 c/iw, B41 c/iw, 1 or more selected from B48 c/iw, B57 c/iw, B68 c/iw, B93 c/iw, B102 c/iw, B108 c/iw, B122 c/iw, BW74 c/iw, BW83 c/iw , preferably 2 or more, more preferably 3 or more, more preferably 5 or more, more preferably 8 or more, still more preferably 10 or more, and particularly preferably all spatial frequencies.

In a particularly preferred embodiment of the invention, R68 c/iw, R79 c/iw, R80 c/iw, R88 c/iw, R109 c/iw, R112 c/iw, R127 c/iw, R133 c/iw, G33 c/iw, G105 c/iw, G115 c/iw, G125 c/iw, G132 c/iw, G43 c/iw, G110 c/iw, G128 c/iw, B36 c/iw, B41 c/iw, Spatial frequency selected from B48 c/iw, B57 c/iw, B68 c/iw, B93 c/iw, B102 c/iw, B108 c/iw, B122 c/iw, BW74 c/iw, BW83 c/iw (If the sign (positive or negative) of the correlation coefficient shown in Table 1 and FIG. 3 is positive, the higher the strength, the more negative the sign (positive or negative) of the correlation coefficient shown in Table 1 and FIG. The lower the strength in the case), the more excellent the smoothness of the skin is determined, and based on the determination, it is estimated that the face expresses a sense of happiness.

In addition, the estimation method of the present invention includes R10 c/iw, R11 c/iw, R13 c/iw,
G8 c/iw, G12 c/iw, G13 c/iw, B9 c/iw, B15 c/iw, BW2 c/iw, BW46 c/iw, BW63 c/iw, BW75 c/iw, BW85 c/iw, 1 or more, preferably 2 or more, more preferably 3 or more, more preferably 5 or more, more preferably 8 or more, still more preferably 10 or more, particularly preferably all spatial frequencies selected from is preferred.

In a particularly preferred embodiment of the invention, R10 c/iw, R11 c/iw, R13 c/iw, G8 c/iw, G12 c/iw, G13 c/iw, B9 c/iw, B15 c/iw, The strength of spatial frequencies selected from BW2 c/iw, BW46 c/iw, BW63 c/iw, BW75 c/iw, and BW85 c/iw is remarkable (the sign of the correlation coefficient shown in Table 1 and FIG. ) is positive, the higher the intensity, and when the sign (positive or negative) of the correlation coefficient shown in Table 1 and FIG. 3 is negative, the intensity is lower). Based on the determination, it is estimated that the face expresses a sense of happiness.

A specific flow of the estimation method of the present invention will be described below.

Further, as a specific aspect of the estimation method of the present invention,
a facial image acquisition step S1 for acquiring facial image data from a subject;
a channel division step S2 of acquiring R (red), G (green), B (blue), and BW (gray) color channel data from the face image data acquired in the face image acquisition step;
a spatial frequency calculation step S3 for calculating the intensity of the spatial frequency from at least one of RGB and BW color channel data;
Applying the calculated spatial frequency intensity to a previously prepared formula or model showing the correlation between the spatial frequency intensity obtained from the face image data and the evaluation value of the face expressing happiness, An evaluation step S4 for evaluating the degree of the face expressing the feeling of happiness of
a display step S5 for displaying the evaluation result in the evaluation step S4;
(See FIG. 27).

Here, in the present invention, it is not always necessary to perform all the above steps S1 to S5. a calculation step S3;
Applying the calculated spatial frequency intensity to a previously prepared formula or model showing the correlation between the spatial frequency intensity obtained from the face image data and the evaluation value of the face expressing happiness, and an evaluation step S4 of evaluating the degree of happiness expressed by the face.

Further, in the present invention, it is preferable that the form includes update of the DB (database) based on the evaluation result in the evaluation step S4 and/or accumulation of the evaluation result in the evaluation step S4 in the DB (database).
In particular, it is preferable that the update/accumulation described above is fed back and evaluated in the evaluation step S4.

In a preferred embodiment of the present invention, the degree of happiness estimated by the present invention can be used as an index of stimulation suitable for the subject.

Specifically, the present invention provides an applying step of applying any stimulus to the subject;
a measuring step of measuring the degree of happiness-appearing faces obtained before the applying step and estimated by the above-described estimation method, and the degree of change in the degree of happiness-appearing faces after the applying step;
a determination step of determining that the stimulus is suitable for the subject when an increase in the degree of happiness is observed on the face, based on the result of the measurement step;
It is also a method of selecting a stimulus suitable for a subject, characterized by having

Here, the determination step is preferably a step of determining that the stimulus is more suitable for the subject as the rate of increase in the degree of happiness manifested on the face increases.

In addition, optional stimuli include travel experience, conversation experience (counseling, communication), art appreciation (music, drama, performance, painting, reading, art), dining experience, sports experience (exercise experience, sports appreciation), and bathing. Experiences (hot springs, saunas), sleep, animal therapy (contact with animals), creative activities, and hobbies can be mentioned.

In particular, in a preferred embodiment of the present invention, the degree of happiness estimated by the present invention can be used as an index when selecting cosmetics.

Specifically, the present invention provides an application step of applying a skin condition improving agent and/or cosmetic means to a subject;
a measuring step of measuring the degree of happiness-appearing faces obtained before the applying step and estimated by the above-described estimation method, and the degree of change in the degree of happiness-appearing faces after the applying step;
a judgment step of judging that the substance and/or cosmetic means are suitable for the subject and have an effect of improving the skin condition, if an increase in the degree of happiness is observed on the face, based on the results of the measurement step; ,
It is also a method for selecting a skin condition improving agent and / or cosmetic means, characterized by having

Here, the determination step is preferably a step of determining that the stimulus is more suitable for the subject as the rate of increase in the degree of happiness manifested on the face increases.

Skin condition improving agents include commercially available compounds (including peptides), known compounds (including peptides), compounds obtained by combinatorial chemistry technology, natural ingredients derived from plants and marine organisms, animal tissue extracts, etc. Substances can be mentioned.

Animal and plant-derived extracts refer not only to animal- or plant-derived extracts per se, but also to collectively include fractions of extracts, purified fractions, extracts or fractions, and solvent-removed purified products.
Plant-derived extracts include wild or grown plants, extracts using those sold as raw materials for herbal medicines, commercially available extracts, and the like.

Cosmetic means include face packs, massage techniques, aromas, and the like.

In addition, the result of the degree of happiness expressed on the face determined by the present invention can also be a useful index in counseling regarding skin care (skin care) and makeup methods.

Here, the preferred embodiment of the method for selecting the skin condition improving agent and/or cosmetic means of the present invention can refer to the above description.

<Screening method for skin condition improving agent>
The present invention is also a screening method for a skin condition improving agent based on the correlation between the intensity of spatial frequency obtained from face image data and the evaluation value of a face expressing happiness.

The present invention
an application step of applying the substance to the skin;
a measuring step of measuring the degree of happiness-appearing faces obtained before the applying step and estimated by the above-described estimation method, and the degree of change in the degree of happiness-appearing faces after the applying step;
A determination step of determining that the higher the rate of increase in the degree of happiness manifestation of the face, the better the skin condition improving agent based on the results of the measurement step;
It can also be a form having

In addition, the present invention provides an application step of applying two or more substances to two or more subjects, respectively;
a measuring step of measuring, after the applying step, the degree of happiness-appearing faces obtained in advance before the applying step and the degree of change in the degree of happiness-appearing faces after the applying step;
A determination step of determining that a skin condition improving agent with a higher rate of increase in the degree of happiness expressed on the face is a superior skin condition improving agent based on the measurement results of the measuring step;
It can also be a form having

In the screening method of the present invention, the skin condition improving agent to be screened is not particularly limited, and examples include commercially available compounds (including peptides), known compounds (including peptides), and compounds obtained by combinatorial chemistry techniques. Groups, natural ingredients from plants and marine organisms, and substances such as animal tissue extracts can be mentioned.

Animal and plant-derived extracts refer not only to animal- or plant-derived extracts per se, but also to collectively include fractions of extracts, purified fractions, extracts or fractions, and solvent-removed purified products.
Plant-derived extracts include wild or grown plants, extracts using those sold as raw materials for herbal medicines, commercially available extracts, and the like.

In the present invention, it is particularly preferable to screen agents containing plant-derived extracts.

In the present invention, two or more, preferably three or more types of candidate substances may be applied.
According to the present invention, it is possible to screen for a skin condition improving agent that exerts a synergistic skin condition improving action when multiple types of candidate substances are applied.

In addition, the screening method of the present invention can also be used as a screening target even in the form of cosmetics.
Here, the cosmetics are preferably those applied to the skin. Moreover, as cosmetics, for example, wash, lotion, and milky lotion can be mentioned preferably.

In the present invention, the duration of the application step is preferably 2 weeks or longer, more preferably 3 weeks or longer, still more preferably 6 weeks or longer, and particularly preferably 10 weeks or longer.
By performing the application step for the above period or longer, the skin condition improving agent can be screened more accurately.

Here, preferred embodiments of the screening method of the present invention can refer to the above description.

<Apparatus for estimating the degree of happiness expressed on the face>
The estimation device of the present invention will be described below with reference to the hardware block diagram (FIG. 28) of the estimation device of the present invention.

The estimating apparatus 1 of the present invention includes face image acquisition means 11 for acquiring face image data of a subject, channel dividing means 12 for dividing the acquired face image data into specific color channels, and color face image data of the subject. spatial frequency calculation means 13 for calculating the intensity of spatial frequencies obtained from data obtained from arbitrary channels alone or in combination among RGB and BW color channel data of the face image data; a storage unit 14 for storing correlation data indicating the correlation between the intensity of happiness and an evaluation value of a face expressing happiness; an evaluation unit 15; and a display unit 16 for displaying the evaluation result of the evaluation unit 15.

The face image acquiring means 11 is means for acquiring a face image obtained by a device for inputting face image data, such as a digital camera, a digital microscope, a digital video or a scanner.

The channel division means 12 divides the color face image data acquired by the face image acquisition means 11 into RGB and BW color channels in order to supply the color face image data acquired by the face image acquisition means 11 to the spatial frequency calculation means 13, which will be described later. (1) to generate grayscale image data (RGB and BW color channel data of a face image) that retains any of the hues.

The spatial frequency calculation means 13 calculates the intensity of the spatial frequency by FFT processing, an expression showing the relationship between the face expressing happiness (the face expressing happiness or the face expressing happiness) and the intensity of the spatial frequency Execute processing such as calculation of a model (correlation data).

The storage unit 14 stores a program necessary for the functioning of the apparatus for estimating the degree of happy faces, and various formulas or models (correlations) necessary for estimating the degree of happy faces. data). The storage unit 14 may further store a database necessary for creating formulas or models.

The evaluation means 15 is means for evaluating the degree of the subject's feeling of happiness on the face based on the intensity of the spatial frequency calculated by the spatial frequency calculation means 13 and the correlation data stored in the storage unit 14. is.

The display means 16 is means for instructing an arbitrary display device to display the evaluation result by the evaluation means 15 .
Here, any display device may be used as long as it can display the estimated result of the degree of happiness manifestation of the face. printers and the like.

In the present invention, it is not always necessary to have the above configuration. For example, spatial frequency calculation means 13 for calculating the intensity of the spatial frequency obtained from at least one of the RGB and BW color channel data of the face image;
The calculated intensity of spatial frequency is applied to a previously prepared formula or model showing the correlation between the intensity of spatial frequency obtained from face image data and a face expressing happiness, so that the subject feels happiness. Evaluation means 15 for evaluating the degree of the face that appears,
It may be in the form of an estimating device comprising:

In addition, the present invention is based on the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face showing the feeling of happiness, from the face image data of the subject. It may be in the form of an estimator that includes estimating the degree of a face.

Here, the preferred embodiment of the estimation device of the present invention can refer to the above description.

<Program for estimating the degree of happiness expressed on the face>
The estimating program of the present invention comprises a computer, other devices, machines, etc., which are configured to have facial image acquisition means 11 for acquiring facial image data of a subject and channel division means 12 for dividing the acquired facial image data into specific color channels. and a spatial frequency calculation means 13 for calculating the intensity of the spatial frequency obtained by combining data obtained from arbitrary channels among the RGB and BW color channel data of the color face image data of the subject. , a storage unit 14 for storing correlation data indicating the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing happiness, an evaluation means 15, and an evaluation result of the evaluation means 15 are displayed. It functions as display means 16 for displaying (see FIG. 28).

In the present invention, it is not necessary to use a program that functions all of the above configurations. or a spatial frequency calculation means 13 for calculating the intensity of the spatial frequency obtained by the combination;
The calculated intensity of spatial frequency is applied to a previously prepared formula or model showing the correlation between the intensity of spatial frequency obtained from face image data and a face expressing happiness, so that the subject feels happiness. Evaluation means 15 for evaluating the degree of the face that appears,
It may be in the form of an estimation program that functions as

Further, the present invention provides a computer,
Based on the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing happiness, the degree of happiness of the subject's face is estimated from the face image data of the subject. It may be in the form of an estimation program for functioning as described above.

Further, the present invention may be such a program recorded in a computer-readable recording medium.

Here, the preferred embodiment of the estimation program of the present invention can refer to the above description.

The test results on which the present invention is based are shown below.

<1> Evaluation of Faces Expressing Happiness (1) Preparation of Image for Evaluation First, face images of 272 women aged 18 to 82 (average age 49.1, SD=16.91) were photographed. Based on the photographed face image, image data of 1024×1024 pixels for use in the test was created. Here, in preparing the images to be used for the test, the hair and ears were removed from the original image, the left and right pupils were fixed at fixed positions in the image, and the size and position of the face were standardized.

(2) Impression evaluation 20 healthy women (5 in their 20s, 5 in their 30s, 5 in their 40s, 5 in their 50s, average age of 39.7, SD = 12.1) Face” was used as an evaluator.
The above-mentioned evaluator evaluated whether the person of the face image prepared in (1) "looks happy" in the following seven grades (see FIG. 1).
In addition, in order to avoid the bias of the evaluator's reaction, the presentation order of the face image was changed for each evaluator and used for the test. An average evaluation value (hereinafter referred to as a score value) of “a face expressing happiness” was calculated for each face image.

Score 1: Not very happy Score 2: Not very happy Score 3: Slightly unhappy Score 4: Neutral Score 5: Slightly happy Score 6: Fairly happy Visible Grade 7: Looks very happy

(3) Relationship Between Score Value and Age The relationship between score value and age is illustrated with the score value set on the vertical axis and the age on the horizontal axis (Fig. 2).
As shown in Figure 2, score values did not decrease with aging.
That is, it was found that there is no correlation between the score value and age.

(4) Relationship between score value and intensity of each spatial frequency (4-1) Spatial frequency analysis of face image First, the color face image subjected to the test in (2) was divided into RGB and BW color channels by an image processing system. , R (red), G (green), B (blue), and BW (gray) (corresponding to face image data in this specification) were obtained. Color channel data that retains specific hues among R (red), G (green), B (blue), and BW (gray) are processed into spatial frequency analysis software manufactured by TheMathWoRk using MATLAB (registered trademark). and performed spatial frequency analysis.

Spatial frequency analysis using MATLAB (registered trademark), R (red), G (green), B (blue), BW (gray) color channel data of the face image subjected to the test of (2) , a power value (spatial frequency intensity) of 1 to 512 cycle/image-width (hereinafter referred to as c/iw) was obtained.

(4-2) Derivation of correlation coefficient between score value and power value of each spatial frequency Score value for each color channel data of R (red), G (green), B (blue), BW (gray) and the correlation coefficient of the power value of each spatial frequency. The results are shown in FIG.
As shown in FIG. 3, a significant power value correlated with the score value could be confirmed.

(4-3) Multiple regression analysis A multiple regression analysis was performed on the score value and the power value of each spatial frequency to confirm the characteristic power value.
Here, the multiple regression analysis was performed by the stepwise method using the score value as the objective variable and the power value of each spatial frequency as the explanatory variable.
Table 1 shows the results. In Table 1, items are selected in which the sign (positive or negative) of the estimated value (hereinafter simply referred to as the estimated value) of the multiple regression analysis result matches the sign (positive or negative) of the correlation coefficient shown in FIG.

Hereinafter, in this embodiment, data obtained from color channel data holding a specific hue among color channel data of any of R (red), G (green), B (blue), and BW (gray) When referring to the spatial frequency range of , it is expressed as "hue spatial frequency range of the divided color channels".
For example, among R (red), G (green), B (blue), and BW (gray) color channel data, the spatial frequency range of the data obtained from the R (red) color channel data is 4 c/iw. is expressed as "R 4 c/iw". (See Table 1 Fig. 5).

R-squared: 0.444658
Degree of freedom adjustment R2: 0.402787
Analysis of variance: F value p<0.0001

Estimation formula:
= -4.433832+(1.0291806*R4)+(2.1284523*R6)+(-1.14673*R10)+(1.3188023*R11)+(3.1612062*R12)+(-2.509092)+(0.74)+ 06471*R24)+(- 1.540866*R29)+(-2.062135*R68)+(-3.977294*R79)+(1.8109804*R80)+(1.5980875*R88)+(-3.278925*R109)+(2.0183678*R112)+(-2.00003 7*R127) +(4.1881341*R133)+(-0.418772*G5)+(1.1394821*G8)+(-2.334774*G12)+(1.634245*G13)+(-2.010401*G22)+(0.8383603*G26)+(1.3526653* G33 )+(-1.684988*G43)+(-2.425363*G105)+(-2.259147*G110)+(4.125416*G115)+(2.454947*G125)+(-2.971531*G128)+(1.9774756*G132)+(- 0.791885*B9)+(-1.018734*B15)+(-2.012453*B28)+(1.8795505*B36)+(1.7686815*B41)+(-1.311655*B47)+(1.355811*B48)+(2.8066131*B5) 7)+ (3.6181894*B68)+(1.8173531*B93)+(-2.622764*B102)+(4.0632072*B108)+(-2.448105*B122)+(-0.649974*BW2)+(-1.1173*BW46)+(-2.5033 34* BW63)+(2.1046171*BW74)+(-2.912047*BW75)+(2.382319*BW83)+(-2.74303*BW85)

For each face image data acquired in (1), the score value is set on the vertical axis, and the estimated value by the multiple regression equation (hereinafter simply referred to as the estimated value) is set on the horizontal axis, and the relationship between the score value and the estimated value is illustrated ( Figure 4).
As shown in FIG. 4, it was found that the higher the score value, the higher the estimated value.
In other words, it was found that the multiple regression equation derived in this example shows the correlation between the score value (the degree of happiness expressed by the face) and the estimated value.

<2> Verification of Estimated Value of Degree of Happiness Appearing on Faces The results derived in <1> were verified.

Verification was carried out as to what kind of features the respective power values at the spatial frequencies shown in Table 1 reflected in the impression of the face.
For the above verification, an image was produced in which the power value at a specific spatial frequency was removed or emphasized, and the image was subjected to evaluation by an expert who specializes in techniques relating to skin conditions.

(1) Verification of what features are reflected in the face image by the power value of the spatial frequency region R4 c/iw

As shown in Table 1, the correlation coefficient between the power value of R4 c/iw and the happy face is -0.00271. Then, the lower the power value of R4 c/iw, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of R4 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 5, it was found that the lower the power value of R4 c/iw, the more happier the face was.

As shown in Fig. 5, in the image with the power value removed (0x processing), the entire cheek, forehead, and chin are dark, and the bridge of the nose and the upper part of the outer corner of the eye (T-shaped eyes and nose) are bright, giving a three-dimensional effect. I found that it looks like

As shown in FIG. 5, in the image in which the power value was emphasized (quintuple processing), the cheeks, the forehead, and the chin were strongly reddish, and it was found that the three-dimensional effect of the cheeks was lost, and the person did not look happy.

In addition, as shown in FIG. 5, in the image in which the power value was emphasized (5 times processing), the bridge of the nose and the upper part of the outer corner of the eye (T-shape of the eyes and nose) were dark, and it was found that the three-dimensional effect was lost and the person did not look happy. .

(2) Verification of what features are reflected in the face image by the power value of the spatial frequency region R6 c/iw

As shown in Table 1, the correlation coefficient between the power value of R6 c/iw and the happy face is 0.1882. Then, the higher the power value of R6 c/iw, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of R6 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 6, it was found that the higher the power value of R6 c/iw, the more happier the face was.

As shown in FIG. 6, in the image from which the power value is removed (0x processing), the raised cheeks, the side of the mouth, the root of the nose, the upper part of the corner of the eye, and the center of the forehead are dark, and the feeling of lift-up is lost. I found that I didn't look happy.

In addition, as shown in Fig. 6, in the image in which the power value is emphasized (5 times processing), redness is strong in the raised part of the cheek, the side of the mouth, the root of the nose, the upper part of the corner of the eye, and the center of the forehead, and the cheek is lifted. I found that I can look happy by doing it.
In addition, it was found that in the image in which the power value was emphasized (quintuple processing), the bridge of the nose looked taller and the person looked happier.
In addition, in the image in which the power value was emphasized (quintuple processing), the area between the eyebrows, the lower part of the inner corner of the eye, and the upper part of the corner of the mouth were dark, and it was found that the contrast gave a three-dimensional effect, which made the person look happy.

(3) Verification of what features are reflected in the face image by the power value of the spatial frequency region R10 c/iw

As shown in Table 1, the correlation coefficient between the power value of R10 c/iw and the happy face is -0.021. Then, the lower the power value of R10 c/iw, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of R10 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 7, it was found that the lower the power value of R10 c/iw, the more happier the face was.

As shown in FIG. 7, in the image from which the power value was removed (0-fold processing), the raised cheeks were bright, and it was found that the person looked happy rather than having a three-dimensional effect.

As shown in FIG. 7, in the image in which the power value was emphasized (5 times processing), it was found that the area under the eyes, the top of the nose root, and the center of the forehead were dark, and the subjects looked less happy than the dark circles under the eyes.
As shown in FIG. 7, in the image in which the power value was emphasized (quintuple processing), the entire cheek was reddish, and it was found that the three-dimensional effect of the cheek was lost and the subject did not look happy.

(4) Verification of what features are reflected in the face image by the power value of the spatial frequency region R11 c/iw

As shown in Table 1, the correlation coefficient between the power value of R11 c/iw and the happy face is 0.0988. Then, the higher the power value of R11 c/iw, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of R11 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 8, it was found that the higher the R11 c/iw power value, the more happier the face was.

As shown in FIG. 8, it was found that the image from which the power value was removed (0-fold processing) lost the contrast and the three-dimensional effect was lost, and the person did not look happy.

In addition, in the image in which the power value was emphasized (quintuple processing), the corners of the eyes, under the lower lip, and under the nose were strongly reddish. In addition, in the image with the power value emphasized (5 times processing), the area between the eyebrows, the upper and lower parts of the inner corner of the eye, the tip of the nose, the lower part of the outer corner of the eye, the center of the lower lip and the chin are dark, and the contrast gives a three-dimensional effect, making it look happy. have understood.

(5) Verification of what features are reflected in the face image by the power value of the spatial frequency region R12 c/iw

As shown in Table 1, the correlation coefficient between the power value of R12 c/iw and the happy face is 0.0146. Then, the higher the power value of R12 c/iw, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of R12 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 9, it was found that the higher the R12 c/iw power value, the more happier the face was.

As shown in FIG. 9, it was found that the image from which the power value was removed (processed by a factor of 0) lost the contrast and the three-dimensional effect was lost, so that the person did not look happy.

As shown in Fig. 9, in the image in which the power value is emphasized (5 times processing), redness is strong under the eyes, between the root of the nose and the eyebrows, under the lower lip, and above the upper lip. I understand.
Moreover, as shown in FIG. 9, it was found that in the image in which the power value was emphasized (quintuple processing), the area around the mouth appeared bright, giving a three-dimensional effect and making the person look happy.

(6) Verification of what features are reflected in the face image by the power value of the spatial frequency region R13 c/iw

As shown in Table 1, the correlation coefficient between the power value of R13 c/iw and the happy face is -0.0574. Then, the lower the power value of R13 c/iw, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of R13 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 10, it was found that the lower the R13 c/iw power value, the more happier the face was.

As shown in FIG. 10, it was found that the image from which the power value was removed (0-fold processing) gave a three-dimensional effect due to the contrast, and that the image looked happier.

As shown in FIG. 10, in the image in which the power value is emphasized (5 times processing), the area under the inner corner of the eye (golgo line), between the eyebrows, under the corner of the eye, and the side of the corner of the mouth are dark. I found that I couldn't see it. In addition, as shown in FIG. 10, in the image in which the power value was emphasized (quintuple processing), the entire cheeks were strongly reddish, and it was found that the three-dimensional effect of the cheeks was lost, and the person did not look happy.

(7) Spatial frequency domain Verification of what features are reflected in the face image by the power values of R24 c/iw and R29 c/iw

As shown in Table 1, the correlation coefficient between the power value of R24 c/iw and the happy face is -0.0426.
The correlation coefficient between the power value of R29 c/iw and the happy face is -0.1171.
Then, the lower the power value, the more the face is evaluated as a face expressing a sense of happiness.

To confirm the above evaluation, images were prepared in which the power values of R24 c/iw and R29 c/iw were removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 11, it was found that the lower the power values of R24 c/iw and R29 c/iw, the more happier the face was.

As shown in FIG. 11, it was found that the image from which the power value was removed (0-fold processing) gave a three-dimensional effect due to the contrast, and that the image looked happier.

In addition, in the image in which the power value was emphasized (quintuple processing), it was found that the area around the mouth was dark and the corners of the mouth appeared to be lowered, which made the subject look unhappy.
In addition, in the image in which the power value was emphasized (quintuple processing), the entire cheeks were strongly reddish, and it was found that the three-dimensional effect of the cheeks was lost, and the person did not look happy.

(8) Spatial frequency range Verification of what features are reflected in facial images

As shown in Table 1,
The correlation coefficient between the power value of R68 c/iw and the happy face is -0.0936.
The correlation coefficient between the power value of R79 c/iw and the happy face is -0.0923.
The correlation coefficient between the power value of R80 c/iw and the happy face is -0.0723.
The correlation coefficient between the power value of R88 c/iw and the happy face is -0.0239.
The correlation coefficient between the power value of R109 c/iw and the happy face is -0.0992.
The correlation coefficient between the power value of R112 c/iw and the happy face is -0.0353.
The correlation coefficient between the power value of R127 c/iw and the happy face is -0.0541.
The correlation coefficient between the power value of R133 c/iw and the happy face is -0.0639.
Then, the lower the power value, the more the face is evaluated as a face expressing a sense of happiness.

For confirmation of the above evaluation, the power values of R68 c/iw, R79 c/iw, R80 c/iw, R88 c/iw, R109 c/iw, R112 c/iw, R127 c/iw, and R133 c/iw Images were prepared after removal (0x processing) or enhancement (10x processing). The results are shown in FIG.
As shown in FIG. 12, the power values of R68 c/iw, R79 c/iw, R80 c/iw, R88 c/iw, R109 c/iw, R112 c/iw, R127 c/iw, and R133 c/iw are low. It was found that the more the person's face was, the more happier the face was.

As shown in FIG. 12, it was found that the image from which the power value was removed (0-fold processing) appeared to be happier than that the contrast appeared smoother.

As shown in FIG. 12, it was found that the image in which the power value was emphasized (10-fold processing) did not look happy because the smoothness of the entire face was lost.

(9) Verification of what features are reflected in the face image by the power value of the spatial frequency domain G5 c/iw

As shown in Table 1, the correlation coefficient between the power value of G5 c/iw and the happy face is -0.1143. Then, the lower the power value of G5 c/iw, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of G5 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 13, it was found that the lower the power value of G5 c/iw, the more happier the face was.

As shown in FIG. 13, it was found that the image from which the power value was removed (0-fold processing) gave a three-dimensional effect due to the contrast, and that the image looked happier.
As shown in FIG. 13, in the image in which the power value is emphasized (5 times processed), green is strong in the area between the eyebrows and the eyes, and under the lower lip, from the nostrils of the cheeks, giving a dark impression of the face, which makes the face look happy. It turns out not.

(10) Verification of what features are reflected in the face image by the power value of the spatial frequency domain G8 c/iw

As shown in Table 1, the correlation coefficient between the power value of G8 c/iw and the happy face is 0.0576. Then, the higher the power value of G8 c/iw, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of G8 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 14, it was found that the higher the G8 c/iw power value, the more happier the face was.

As shown in FIG. 14, in the image from which the power value was removed (0-fold processing), the area between the eyebrows, the area below the eye, the root of the nose, the tip of the nose, and the corner of the mouth are dark, and the dark circles under the eye and the bridge of the nose appear low. have understood.

As shown in Fig. 14, in the image in which the power value is emphasized (5 times processing), green is strong in the upper and lower corners of the eyes, under the inner corners of the eyes, between the eyebrows and the root of the nose, under the nose, under the lower lip, and on the outside of the cheeks, and the contrast is lifted. I found that I looked happier when I looked up.

(11) Verification of what features are reflected in the face image by the power value of the spatial frequency domain G12 c/iw

As shown in Table 1, the correlation coefficient between the power value of G12 c/iw and the happy face is -0.0028. Then, the lower the power value of G12 c/iw, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of G12 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 15, it was found that the lower the G12 c/iw power value, the more happier the face was.

As shown in FIG. 15, it was found that the image from which the power value was removed (0-fold processing) gave a three-dimensional effect due to the contrast, and that the image looked happier.

As shown in FIG. 15, in the image in which the power value is emphasized (5 times processing), green is strong under the eyes, above the inner corners of the eyes, between the root of the nose and the eyebrows, above the upper lip, below the lower lip, and above the outer corners of the eyes. , It turned out that the impression around the mouth looks bad, so it does not look happy.

(12) Verification of what features are reflected in the face image by the power value of the spatial frequency domain G13 c/iw

As shown in Table 1, the correlation coefficient between the power value of G13 c/iw and the happy face is -0.0692. Then, the lower the power value of G13 c/iw, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of G13 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 16, it was found that the lower the G13 c/iw power value, the more happier the face was.

As shown in FIG. 16, it was found that the image from which the power value was removed (0-fold processing) gave a three-dimensional effect due to the contrast, and that the image looked happier.

As shown in FIG. 16, in the image in which the power value is emphasized (quintuple processing), green is strong under the eyes, above the inner corners of the eyes, on the side of the outer corners of the eyes, on the tip of the nose, above the upper lip, below the lower lip, at the corners of the mouth, and on the chin. It was found that the cheeks, around the eyes, and between the lower lip and chin were darker and lacked three-dimensionality.

(13) Verification of what features are reflected in the face image by the power values of the spatial frequency regions G22 c/iw and G26 c/iw

As shown in Table 1, the correlation coefficient between the power value of G22 c/iw and the happy face is -0.1686.
As shown in Table 1, the correlation coefficient between the power value of G26 c/iw and the happy face is -0.0265.
Then, the lower the power value, the more the face is evaluated as a face expressing a sense of happiness.

To confirm the above evaluation, images were prepared in which the power values of G22 c/iw and G26 c/iw were removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 17, it was found that the lower the power values of G22 c/iw and G26 c/iw, the more happier the face was.

As shown in FIG. 17, it was found that the image from which the power value was removed (0-fold processing) gave a three-dimensional effect due to the contrast, and that the image looked happier.

As shown in FIG. 17, in the image in which the power value is emphasized (quintuple processing), the entire face (large width) is green, and the entire face (small width) is dark. I found that I couldn't see it.

(14) Spatial frequency domain The power values of G33 c/iw, G105 c/iw, G115 c/iw, G125 c/iw, and G132 c/iw reflect what features in the facial image. inspection

As shown in Table 1, the correlation coefficient between the power value of G33 c/iw and the happy face is 0.0622.
As shown in Table 1, the correlation coefficient between the power value of G105 c/iw and the face expressing happiness is 0.0018.
As shown in Table 1, the correlation coefficient between the power value of G115 c/iw and the happy face is 0.0464.
As shown in Table 1, the correlation coefficient between the power value of G125 c/iw and the happy face is 0.0716.
As shown in Table 1, the correlation coefficient between the power value of G132 c/iw and the face expressing happiness is 0.0011.
Then, the higher the power value, the more the face is evaluated as a face expressing a sense of happiness.

For confirmation of the above evaluation, images with the power values of G33 c/iw, G105 c/iw, G115 c/iw, G125 c/iw, and G132 c/iw removed (0x processing) or enhanced (10x processing) prepared. The results are shown in FIG.
As shown in FIG. 18, it was found that the higher the power values of G33 c/iw, G105 c/iw, G115 c/iw, G125 c/iw, and G132 c/iw, the more happier the face was.

As shown in FIG. 18, it was found that the image from which the power value was removed (0-fold processing) did not look happy because the smoothness was lost due to the contrast.

As shown in FIG. 18, in the image in which the power value is emphasized (10x processing), the entire face (small width) is green, and the entire face (small width) is dark. I found that I looked happier than I looked.

(15) Verification of what features are reflected in the face image by the power values of the spatial frequency regions G43 c/iw, G110 c/iw, and G128 c/iw

As shown in Table 1, the correlation coefficient between the power value of G43 c/iw and the face expressing happiness is -0.0315.
As shown in Table 1, the correlation coefficient between the power value of G110 c/iw and the face expressing happiness is −0.0004.
As shown in Table 1, the correlation coefficient between the power value of G128 c/iw and the happy face is -0.0568.
Then, the lower the power value, the more the face is evaluated as a face expressing a sense of happiness.

To confirm the above evaluation, images were prepared by removing (0-fold processing) or enhancing (10-fold processing) the power values of G43 c/iw, G110 c/iw, and G128 c/iw. The results are shown in FIG.
As shown in FIG. 19, it was found that the lower the power values of G43 c/iw, G110 c/iw, and G128 c/iw, the more happier the face was.

As shown in FIG. 19, it was found that the image from which the power value was removed (0-fold processing) looked happier than the smoother contrast.

As shown in FIG. 19, in the image in which the power value is emphasized (10-fold processing), the entire face (small width) is green, and the entire face (small width) is dark, and smoothness is lost due to the effect of contrast. I found that I didn't look happy.

(16) Verification of what features are reflected in the face image by the power value of the spatial frequency domain B9 c/iw

As shown in Table 1, the correlation coefficient between the B9 c/iw power value and the happy face is 0.1055. Then, the higher the power value of B9 c/iw, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of B9 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 20, it was found that the lower the power value of B9 c/iw, the more happier the face was.

As shown in FIG. 20, it was found that the image from which the power value was removed (0-fold processing) did not look happy because the three-dimensional effect was lost due to the effect of the contrast.

As shown in Fig. 20, in the image in which the power value is emphasized (5 times processing), if there is a strong bluish tinge in the raised part of the cheeks, the upper and lower corners of the eyes, the tip of the nose, and the corners of the mouth, the cheeks look lifted and look happier. I found it visible.
As shown in FIG. 20, it was found that in the image in which the power value was emphasized (quintuple processing), the impression around the eyes looked better, and the subject looked happier.
As shown in FIG. 20, it was found that in the image in which the power value was emphasized (quintuple processing), the corners of the mouth seemed to rise and the person looked happy.
As shown in FIG. 20, in the image in which the power value is emphasized (5 times processing), the cheeks are lifted up by the contrast when the area around the raised cheeks, the area above the eyes, the bridge of the nose, the area below the nose, and the area below the lower lip is dark. I found that I looked happier than I looked.
As shown in FIG. 20, it was found that in the image in which the power value was emphasized (quintuple processing), the impression around the eyes and mouth looked better, and the person looked happier.

(17) Verification of what features are reflected in the face image by the power value of the spatial frequency domain B15 c/iw

As shown in Table 1, the correlation coefficient between the B15 c/iw power value and the happy face is -0.1746. Then, the higher the B15 c/iw power value, the more the face is evaluated as having a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of B15 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 21, it was found that the lower the B15 c/iw power value, the more happier the face was.

As shown in FIG. 21, it was found that the image from which the power value was removed (0-fold processing) gave a three-dimensional effect due to the contrast, and that the image looked happier.

As shown in FIG. 21, in the image in which the power value is emphasized (quintuple processing), there is strong green under the eyes, above the inner corners of the eyes, on the sides of the outer corners of the eyes, at the tip of the nose, on the upper and upper lips, below the lower lips, at the corners of the mouth, and on the chin, and around the mouth. It turned out that it doesn't look happy than it looks bad.
As shown in FIG. 21, in the image in which the power value was emphasized (quintuple processing), the cheeks, around the eyes, and between the lower lip and the chin were dark, and it was found that the three-dimensional effect was lost and the person did not look happy.

(18) Verification of what features are reflected in the face image by the power values of the spatial frequency regions B28 c/iw and B47 c/iw

As shown in Table 1, the correlation coefficient between the B28 c/iw power value and the happy face is -0.1054.
As shown in Table 1, the correlation coefficient between the B47 c/iw power value and the happy face is -0.0493.
Then, the lower the power value, the more the face is evaluated as a face expressing a sense of happiness.

To confirm the above evaluation, images were prepared in which the power values of B28 c/iw and B47 c/iw were removed (0-fold processing) or emphasized (10-fold processing). The results are shown in FIG.
As shown in FIG. 22, it was found that the lower the power values of B28 c/iw and B47 c/iw, the more happier the face was.

As shown in FIG. 22, it was found that the image from which the power value was removed (0-fold processing) gave a three-dimensional effect due to the contrast, and that the image looked happier.

As shown in FIG. 22, in the image in which the power value is emphasized (10x processing), the entire face (large width) has a strong bluish tint, or the entire face (large width) is dark, and the three-dimensional effect is lost. I found that I couldn't see it.

(19) Spatial frequency range B36 c/iw, B41 c/iw, B48 c/iw, B57 c/iw, B68 c/iw, B93 c/iw, B102 c/iw, B108 c/iw, B122 c/iw Verification of what features are reflected in the face image by the power value of

As shown in Table 1, the correlation coefficient between the B36 c/iw power value and the happy face is 0.0677.
As shown in Table 1, the correlation coefficient between the power value of B41 c/iw and the happy face is 0.122.
As shown in Table 1, the correlation coefficient between the B48 c/iw power value and the happy face is 0.1056.
As shown in Table 1, the correlation coefficient between the B57 c/iw power value and the happy face is 0.1093.
As shown in Table 1, the correlation coefficient between the B68 c/iw power value and the happy face is 0.0463.
As shown in Table 1, the correlation coefficient between the B93 c/iw power value and the happy face is 0.0469.
As shown in Table 1, the correlation coefficient between the power value of B102 c/iw and the face expressing happiness is 0.0567.
As shown in Table 1, the correlation coefficient between the B108 c/iw power value and the happy face is 0.0775.
As shown in Table 1, the correlation coefficient between the B122 c/iw power value and the happy face is 0.0591.
Then, the higher the power value, the more the face is evaluated as a face expressing a sense of happiness.

To confirm the above assessment, B36 c/iw, B41 c/iw, B48 c/iw, B57 c/iw, B68 c/iw, B93 c/iw, B102 c/iw, B108 c/iw, B122 c/ An image was prepared in which the power value of iw was removed (0-fold processing) or emphasized (10-fold processing). The results are shown in FIG.
As shown in FIG. 23, it was found that the higher the power values of B28 c/iw and B47 c/iw, the more happier the face was.

As shown in FIG. 23, it was found that the image from which the power value was removed (0-fold processing) did not look happy because the smoothness was lost due to the contrast.

As shown in FIG. 23, in the image in which the power value is emphasized (10-fold processing), the entire face (small width) has a strong bluish tint, and the entire face (small width) is dark. I found that I looked happier than I looked.

(20) Verification of what features the power value of the spatial frequency domain BW2 c/iw reflects in the facial image

As shown in Table 1, the correlation coefficient between the B28 c/iw power value and the happy face is -0.1028.
Then, the lower the power value, the more the face is evaluated as a face expressing a sense of happiness.

To confirm the above evaluation, an image was prepared in which the power value of BW2 c/iw was removed (0-fold processing) or emphasized (5-fold processing). The results are shown in FIG.
As shown in FIG. 24, it was found that the lower the power value of BW2 c/iw, the more happier the face was.

As shown in FIG. 24, it was found that the image from which the power value was removed (0-fold processing) gave a three-dimensional effect due to the contrast, and that the image looked happier.

As shown in FIG. 24, in the image in which the power value is emphasized (quintuple processing), the central part of the face is bright, and the lack of stereoscopic effect due to the contrast makes it difficult to look happy.

(21) Verification of what features are reflected in the face image by the power values of the spatial frequency regions BW46 c/iw, BW63 c/iw, BW75 c/iw, and BW85 c/iw

As shown in Table 1, the correlation coefficient between the BW46 c/iw power value and the happy face is -0.0789.
As shown in Table 1, the correlation coefficient between the power value of BW63 c/iw and the happy face is -0.0673.
As shown in Table 1, the correlation coefficient between the power value of BW75 c/iw and the face expressing happiness is −0.0769.
As shown in Table 1, the correlation coefficient between the power value of BW85 c/iw and the happy face is -0.0366.
Then, the lower the power value, the more the face is evaluated as a face expressing a sense of happiness.

To confirm the above evaluation, images were prepared by removing (0-fold processing) or enhancing (10-fold processing) the power values of BW46 c/iw, BW63 c/iw, BW75 c/iw, and BW85 c/iw. The results are shown in FIG.
As shown in FIG. 25, it was found that the lower the BW2 c/iw power value, the more happier the face was.

As shown in FIG. 25, it was found that the image from which the power value was removed (0-fold processing) gave a three-dimensional effect due to the contrast, and that the image looked happier.

As shown in FIG. 25, in the image in which the power value is emphasized (10-fold processing), the whole face (large width) is bright and the whole face (large width) is dark, and the three-dimensional effect is lost. I found that I couldn't see it.

(22) Verification of what features are reflected in the facial image by the power values of the spatial frequency regions BW74 c/iw and BW83 c/iw

As shown in Table 1, the correlation coefficient between the power value of BW74 c/iw and the face expressing happiness is 0.0403.
As shown in Table 1, the correlation coefficient between the BW83 c/iw power value and the happy face is 0.0258.
Then, the higher the power value, the more the face is evaluated as a face expressing a sense of happiness.

To confirm the above evaluation, images were prepared in which the power values of BW74 c/iw and BW83 c/iw were removed (0-fold processing) or emphasized (10-fold processing). The results are shown in FIG.
As shown in FIG. 26, it was found that the higher the power value of BW2 c/iw, the more happier the face was.

As shown in FIG. 26, it was found that the image from which the power value was removed (0-fold processing) did not look happy because the smoothness was lost due to the influence of the contrast.

As shown in FIG. 26, in the image in which the power value is emphasized (10 times processing), the whole face (small width) looks smoother due to the effect of the contrast between brightness and darkness, which makes the face look happier. have understood.

(23) Results and Discussion Table 2 shows the results of the verifications of (1) to (22) above.

From the results of this embodiment, it was found that the expression of happiness of the subject was found from the face image data of the subject based on the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face showing the happiness. We found that it is possible to estimate the degree of facial recognition.

In particular, from the results of this example, R4 c/iw, R6 c/iw, R10 c/iw, R11 c/iw, R24 c/iw, R29 c/iw, G5 c/iw, G8 c/iw, G13 c/iw, G22 c/iw, G26 c/iw, B9 c/iw, B15 c/iw, B28 c/iw, B47 c/iw, BW2 c/iw, BW46 c/iw, BW63 c/iw, BW75 The intensity of the spatial frequency selected from c/iw and BW85 c/iw is remarkable (when the sign (positive or negative) of the correlation coefficient shown in Table 1 and FIG. 3 is positive, the higher the intensity, the higher the intensity is. If the sign (positive or negative) of the correlation coefficient shown in is negative, the intensity is low), it is determined that the degree of cheek lift-up is excellent, and based on the determination, it can be estimated that the face shows a sense of happiness. I found out.

Further, from the results of this example, R4 c/iw, R11 c/iw, R12 c/iw, R13 c/iw, R24 c/iw, R29 c/iw, G5 c/iw, G8 c/iw, G12 Intensity of spatial frequency selected from c/iw, G13 c/iw, B9 c/iw, B15 c/iw, BW2 c/iw, BW46 c/iw, BW63 c/iw, BW75 c/iw, BW85 c/iw (If the sign (positive or negative) of the correlation coefficient shown in Table 1 and FIG. 3 is positive, the higher the intensity, if the sign (positive or negative) of the correlation coefficient shown in Table 1 and It was found that the lower the intensity), the more excellent the three-dimensional effect around the mouth is, and based on this determination, it can be estimated that the face expresses a sense of happiness.

Further, from the results of this example, R10 c/iw, G5 c/iw, G8 c/iw, G12 c/iw, G22 c/iw, G26 c/iw, B9 c/iw, B15 c/iw, B28 The intensity of the spatial frequency selected from c/iw and B47 c/iw is remarkable (when the sign (positive or negative) of the correlation coefficient shown in Table 1 and FIG. 3 is positive, the higher the intensity, the higher the intensity is. If the sign (positive or negative) of the correlation coefficient shown in is negative, the strength is low), it is judged that the uniformity of the skin color is excellent, and based on this judgment, it can be estimated that the face expresses a feeling of happiness. rice field.

Further, from the results of this example, R68 c/iw, R79 c/iw, R80 c/iw, R88 c/iw, R109 c/iw, R112 c/iw, R127 c/iw, R133 c/iw, G33 c/iw, G105 c/iw, G115 c/iw, G125 c/iw, G132 c/iw, G43 c/iw, G110 c/iw, G128 c/iw, B36 c/iw, B41 c/iw, B48 of spatial frequencies selected from c/iw, B57 c/iw, B68 c/iw, B93 c/iw, B102 c/iw, B108 c/iw, B122 c/iw, BW74 c/iw, and BW83 c/iw The intensity is remarkable (when the sign (positive or negative) of the correlation coefficient shown in Table 1 and FIG. 3 is positive, the higher the intensity, the more The lower the intensity), the smoother the skin is judged to be, and based on this judgment, it can be estimated that the face expresses a sense of happiness.

The technology for estimating the degree of happiness expressed on a face according to the present invention can be used in evaluating the effects of esthetics, skin care, make-up, health food, cosmetic surgery, etc., or in giving advice or counseling using the evaluation results.

Claims (14)

Based on the correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing happiness, the degree of happiness of the subject's face is estimated from the face image data of the subject. Estimation methods, including a spatial frequency calculation step of calculating the strength of the spatial frequency obtained by combining data obtained from arbitrary channels among the RGB and BW color channel data of the color face image data of the face image;
Applying the calculated spatial frequency intensity to a previously prepared formula or model showing the correlation between the spatial frequency intensity obtained from the face image data and the evaluation value of the face expressing happiness, an evaluation process for evaluating the degree of happiness of the face,
The estimation method according to claim 1, comprising: 3. The estimation method according to claim 2, wherein said formula is a multiple regression formula obtained by multiple regression analysis or a prediction formula obtained by PLS. In a preferred embodiment of the present invention, the degree of the face expressing happiness is the degree of lift-up of the cheeks, the stereoscopic effect around the mouth, the uniformity of the skin color, the smoothness of the skin, and the brightness around the eyes. The estimation method according to any one of claims 1 to 3, including at least one degree. The estimation method according to any one of claims 1 to 4, wherein the face image data is face image data of a part of the face or the whole face. 6. The facial image data is color facial image data, and the spatial frequency is a spatial frequency obtained from at least one of RGB and BW color channel data of the color facial image data. The estimation method according to item 1. 7. The estimation method according to claim 6, wherein said spatial frequency is a spatial frequency obtained from a plurality of color channel data among RGB and BW color channel data of a face image. The estimation method according to any one of claims 1 to 7, wherein the spatial frequency range (cycle/image-width) of said spatial frequency is 0 to 512 cycle/image-width for an image having an image length of 1024 pixels. An application step of applying any stimulus (except for medical treatment) to the subject;
The degree of the face expressing happiness estimated by the estimation method according to any one of claims 1 to 8, obtained before the applying step, and the degree of the face expressing happiness after the applying step. a measuring step of measuring the degree of change;
a determination step of determining that the stimulus is suitable for the subject when an increase in the degree of happiness is observed on the face, based on the result of the measurement step;
A method for selecting a stimulus suitable for a subject, characterized by comprising: A screening method for a skin condition improving agent based on the correlation between the intensity of spatial frequency obtained from face image data and the evaluation value of a face expressing happiness. an application step of applying the substance to the skin;
The degree of the face expressing happiness estimated by the estimation method according to any one of claims 1 to 8, obtained before the applying step, and the degree of the face expressing happiness after the applying step. a measuring step of measuring the degree of change;
A determination step of determining that a skin condition improving agent with a higher rate of increase in the degree of happiness expressed on the face is a superior skin condition improving agent based on the results of the measuring step;
The screening method according to claim 10, characterized by having Spatial frequency calculation means for calculating the strength of spatial frequencies obtained by combining data obtained from arbitrary channels among RGB and BW color channel data of color face image data, and
The calculated intensity of spatial frequency is applied to a formula or model prepared in advance showing the correlation between the intensity of spatial frequency obtained from face image data and the face on which feeling of happiness is expressed, and the intensity of feeling of happiness of the subject is calculated. an evaluation means for evaluating the degree of expression of the face;
An estimator, comprising: the computer,
Spatial frequency calculation means for calculating the strength of spatial frequencies obtained by combining data obtained from arbitrary channels among RGB and BW color channel data of color face image data, and
The calculated intensity of spatial frequency is applied to a formula or model prepared in advance showing the correlation between the intensity of spatial frequency obtained from face image data and the face on which feeling of happiness is expressed, and the intensity of feeling of happiness of the subject is calculated. an evaluation means for evaluating the degree of expression of the face;
An estimation program to function as a storage unit that stores, as input information, a correlation between the intensity of the spatial frequency obtained from the face image data and the evaluation value of the face expressing happiness;
a learning unit that generates a learned model in which a variable for estimating the degree of the subject's face showing a sense of happiness is machine-learned from the subject's face image data using the correlation as teacher data;
A learning device comprising:

Images